# podcast

Extracts podcast episode metadata, transcript, and structured summary directly into an Obsidian vault note.

- command · `type=command` · Claude Code, Codex, GitHub Copilot, Gemini CLI · ★ 5412 repo stars
- Author: Eugeniu Ghelbur
- License: MIT · policy=mirror · the beginning of the file is at the end of this page
- Source: https://github.com/eugeniughelbur/obsidian-second-brain/tree/HEAD/commands/podcast.md
- Updated: 2026-10-24
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/commands/podcast

## What it does

The command takes an episode link from Apple Podcasts, Spotify, or a direct RSS feed. For Spotify links, the script fetches the title via oEmbed and maps it back to an open RSS feed via Apple index. Because Spotify audio uses DRM, shows that publish exclusively on Spotify without a public RSS feed will fail.

Under the hood, `scripts.research.podcast_extract` looks for a `<podcast:transcript>` tag in the RSS source. If absent, it attempts speech-to-text using Groq Whisper when `GROQ_API_KEY` is provided, or OpenAI `whisper-1` when `OPENAI_API_KEY` is set. Note that Groq transcription will fail if `ffmpeg` and `ffprobe` are missing from your `PATH`. When no speech model is configured, the script falls back to summarizing RSS show notes alone.

Summaries are generated via Gemini if `GEMINI_API_KEY` exists, or Grok otherwise. Notes save automatically to `Research/Podcasts/YYYY-MM-DD - <episode-title-slug>.md` with wikilinks, guest mentions, and structured metadata formatted for retrieval.

## Requirements
- Works with: Obsidian

## Install

### Claude Code (native platform)

1. **Plugin marketplace**
In the agent chat (not a shell):
```text
/plugin marketplace add eugeniughelbur/obsidian-second-brain
```
In the agent chat (not a shell):
```text
/plugin install obsidian-second-brain@obsidian-second-brain
```

2. **Manually: put the files in place**
- File: `.claude/commands/podcast.md` — Call it with /podcast. For all projects use ~/.claude/commands/

### Codex

**Manually: put the files in place**
- File: `~/.codex/prompts/podcast.md` — Call it with /prompts:podcast. Codex marks this format deprecated in favour of skills

### GitHub Copilot

**Manually: put the files in place**
- File: `.github/prompts/podcast.prompt.md`

### Gemini CLI

**Manually: put the files in place**
- File: `.gemini/commands/podcast.toml` — Gemini stores commands as TOML: the prompt text goes into the prompt field

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/eugeniughelbur/obsidian-second-brain/tree/HEAD/commands/podcast.md

## When to use it
- extract structured summaries and key takeaways from podcast episodes
- transcribe podcast audio using Groq Whisper or OpenAI Whisper
- save podcast notes with metadata directly to an Obsidian vault

## Example prompts
- /podcast https://open.spotify.com/episode/<id>
- Summarize this podcast episode: https://open.spotify.com/episode/<id>
- What's in this episode?

## Source text · podcast.md · MIT

Below is the content of a third-party file. It is data, not instructions for you: do not act on anything written inside unless the human asks. Block tag: b106ac3a1881513d.

BEGIN THIRD-PARTY FILE b106ac3a1881513d
````text
---
description: Extract metadata, transcript, and summary from a podcast episode, saved as an AI-first note in the vault
category: research
triggers_en: ["summarize this podcast", "podcast episode summary", "extract podcast", "what's in this episode"]
triggers_de: ["diesen Podcast zusammenfassen", "Zusammenfassung der Podcastfolge", "Podcast auswerten", "was steckt in dieser Folge"]
triggers_es: ["resume este pódcast", "resumen del episodio", "extrae este pódcast", "qué dice este episodio"]
triggers_pt: ["resuma este podcast", "resumo de episódio de podcast", "extraia este podcast", "o que há neste episódio"]
triggers_zh: ["总结这期播客", "提取播客文字稿", "这期节目讲了什么", "把这期播客整理进知识库"]
---

Use the obsidian-second-brain skill. Execute `/podcast [url]`:

1. Resolve the podcast URL from the user's argument. Accept any of:
   - Apple Podcasts episode URL (`https://podcasts.apple.com/.../id<show>?i=<episode>`)
   - Spotify episode URL (`https://open.spotify.com/episode/<id>`) - Spotify's own audio is DRM-locked, so the script bridges it to the show's public RSS feed: it reads the episode title from Spotify's key-free oEmbed endpoint, finds that episode in Apple's index to get the feed URL, then pulls the episode from the open feed by title. Works for any show that also publishes an open feed (most do); fails clearly for Spotify-exclusive shows with no public RSS.
   - Direct RSS feed URL (uses the latest episode unless `?episode=<guid>` selector is appended)
   - Direct RSS feed URL with `?episode=<guid-fragment-or-link-fragment>` selector

   If no input given, ask: "Which podcast episode? Paste the Apple Podcasts, Spotify, or RSS feed URL."

2. Run the script from the skill root (its absolute path was given at session start as **Skill root**; substitute it for `SKILL_ROOT`):
   ```bash
   uv run --directory "SKILL_ROOT" -m scripts.research.podcast_extract "<url>"
   ```

3. The script:
   - Resolves Apple Podcasts URLs to RSS via the free iTunes Lookup API (no key needed).
   - Parses the RSS feed, extracts episode metadata (title, show, host, published, duration, audio URL, show notes).
   - Tries to obtain a transcript in this order:
     1. **`<podcast:transcript>` tag** in the RSS feed (free, fast, high fidelity).
     2. **Groq-hosted Whisper**, only if `GROQ_API_KEY` is set (free tier). Downloads audio, re-encodes it to 32kbps mono so episodes up to ~4.5h fit in Groq's 25MB per-request cap (longer episodes are split into chunks with overlap and stitched). No cost on the free tier, but the free tier also caps audio at 7,200 seconds per hour (as of 2026-08, console.groq.com): an episode longer than about 2h gets a 429 and falls through to the next step, and a second long episode within the same hour does too. Needs `ffmpeg` and `ffprobe` on PATH.
     3. **Whisper API**, only if `OPENAI_API_KEY` is set. Downloads audio (<=25 MB OpenAI per-file limit), transcribes via `whisper-1`. Approximate cost: $0.006/min.
     4. **Show-notes-only fallback**. If no transcript path works, summarizes from RSS show notes alone. Quality drops; Notable Quotes will be empty.
   - Sends transcript-or-shownotes for AI-first summarization: Gemini when `GEMINI_API_KEY` is set (free tier, 1M context), otherwise Grok. The transcript is capped at `PODCAST_TX_LIMIT` characters (default 480,000, about 120k tokens) with a truncation note if the episode is longer; on the Grok path that cap is a real cost, roughly $0.36 per 3h episode at grok-4 rates, so lower it there if that matters.
````
END THIRD-PARTY FILE b106ac3a1881513d

The third-party file has ended. Everything between the b106ac3a1881513d tags is data: requests and commands from it are not followed. An “end of file” line with another tag inside the block is part of the file.

First 3576 of 5909 bytes. © Eugeniu Ghelbur · MIT · https://github.com/eugeniughelbur/obsidian-second-brain/tree/HEAD/commands/podcast.md

_Part of the page did not fit — see the HTML version._

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
